Module 6

Question 1

Cohort Longitudinal Studies

Answer 1

• observational
• researcher follows a population (can be divided into subgroups)
• attempting to make inference about cause-effect relationships through observation

Question 2

Comparison Groups

Answer 2

1. Internal Comparison - one cohort - compared within the group
2. External Comparison - subgroups - compared between
3. Comparison with general population rates

Question 3

Cohort Studies Advantages

Answer 3

1. can work out incidence rate and risk
2. can look at cause-effect relationships
3. good when exposure is rare
4. minimizes selection and information bias

Question 4

Cohort Studies Disadvantages

Answer 4

1. Losses to follow up
2. large sample size needed
3. expensive
4. ethics
5. ineffective for rare diseases
6. long time to complete

Question 5

Chi-Square Test

Answer 5

See if there is a significant relationship between 2 categorical variables.
Assumption: expected frequency in each cell > 5

Question 6

Fisher’s Exact test

Answer 6

Same as Chi-Square but used if the expected frequency in one or more cells is <5

Question 7

Odds Ratio

Answer 7

Measures the strength of association between an exposure and an outcome

• OR = 1 - exposure does not effect odds of outcome
• OR >1 - exposure increases odds of outcome
• OR <1 - exposure decreases odds of outcome

Question 8

Confidence Interval of Odds Ratio

Answer 8

If CI contains 1 - relationship likely to be insignificant

If CI does not contain 1 - relationship likely to be significant

Question 9

Logistic Regression

Answer 9

a regression with an outcome variable that is categorical and IV’s that are continuous/categorical/mixed

Question 10

Logistic Regression Assumptions

Answer 10

1. Ratio of cases to variables (ie. large sample size)
2. Regression equation should have linear relationship with logit form of the outcome
3. Absence of multicollinearity and outliers
4. Independence of residuals

Question 11

Logistic Regression Models

Answer 11

Binary LR (if dichotomous outcomes)
Multinomial LR (if polychotomous outcomes)
Ordinal LR (if ordered outcome)

Question 12

Hosmer-Lemeshow goodness of fit

Answer 12

used to examine whether LR model fits the sample data
P < 0.05 - poor fit
P > 0.05 - good fit

Question 13

ROC (receiver operating characteristic) curve

Answer 13

used to measure predictive accuracy of the fitted model
0.5 = no predictive power
1 = perfect predictive power